System and Method for Predicting Customer Lifetime Value Using Two-Stage Machine Learning

2. The method of claim 1, wherein the first period of time is earlier than the second period of time and the third period of time.

3. The method of claim 1, wherein the classifier is performed using Random Forrest, and the regressor is performed using Gradient Boosting Decision Tree (GBDT).

4. The method of claim 1, further comprising sending, by the computing device, a promotion package to a predefined percentage of the fourth plurality of customers having the lowest CLVs.

5. The method of claim 4, wherein the predefined percentage is 10%.

6. The method of claim 1, further comprising presenting an advertisement to a portion of the fourth plurality of customers having a predefined range of the CLVs.

7. The method of claim 6, wherein the predefined range is top 20% of CLVs when the advertisement is a luxury brand, and the predefined range is 50%-55% of CLVs when the advertisement is a brand targeting youth.

8. The method of claim 1, wherein the customer feature data is prepared from transaction log, browse log, click log, cart log, demographic information, and indirect features of the customers.

9. The method of claim 8, wherein the preparation comprises aggregation and normalization of spendings from the transaction log, aggregation of a number of transactions from the transaction log, a number of browses from the browse log, a number of clicks from the click log, a number of add-to-cart actions from the cart log, and encoding of shipping address, gender, and age from the demographic information.

10. The method of claim 9, wherein the aggregation comprises summation of the number of transactions, the number of browses, the number of clicks, and the number of add-to-cart actions in a time of a month, a season, and a year.

12. The system of claim 11, wherein the first period of time is earlier than the second period of time and the third period of time.

13. The system of claim 11, wherein the classifier is performed using Random Forrest, and the regressor is performed using Gradient Boosting Decision Tree (GBDT).

14. The system of claim 11, wherein the computer executable code is further configured to send a promotion package to 10% of the fourth plurality of customers having the lowest CLVs.

15. The system of claim 11, wherein the computer executable code is further configured to present an advertisement of a luxury brand to a portion of the fourth plurality of customers that have top 20% of the CLVs, or present an advertisement targeting youth to a portion of the fourth plurality of customers having 50%-55% of the CLVs.

16. The system of claim 11, wherein the customer feature data is prepared from transaction log, browse log, click log, cart log, demographic information, and indirect features of the customers, and the preparation comprises aggregation and normalization of spendings from the transaction log, aggregation of a number of transactions from the transaction log, a number of browses from the browse log, a number of clicks from the click log, a number of add-to-cart actions from the cart log, and encoding of shipping address, gender, and age from the demographic information.

17. The system of claim 16, wherein the aggregation comprises summation of the number of transactions, the number of browses, the number of clicks, and the number of add-to-cart actions in a time of a month, a season, and a year.

19. The non-transitory computer readable medium of claim 18, wherein the classifier is performed using Random Forrest, and the regressor is performed using Gradient Boosting Decision Tree (GBDT).